Curran’s three critiques

Curran raises three objections. The first concerns novelty. Given the close resemblance between his ‘representation as intervention’ and our ‘representational interventions’, he questions whether our framing offers something substantively new, and asks that the two be clearly distinguished. He notes that Michel Callon, who we draw on, sets aside ‘the representational idiom’, leading Curran to suggest that our genealogy may need to ground itself in other theoretical traditions.

The second objection concerns the status of the framework. We present DARN as a heuristic that makes no ontological claims, yet Curran identifies places where we appear to make them: that devices and networks ‘undoubtedly possess agency’, that ‘only representations can be performative’, and that performativity attaches to the effects of representations rather than the actions of agents. He argues that the choice of categories, and the inclusion of networks in particular, encodes substantive commitments we neither acknowledge nor defend, and therefore that our heuristic framing conceals rather than avoids ontology.

Curran’s third objection, the one he critiques hardest and most eloquently, concerns representation itself. We use the term throughout without theorizing it. For Curran, a representation is fundamentally a relation of ‘aboutness’, bound up with human intentionality and the distinctive reflexivity of conscious beings, as captured in Hacking’s (Reference Hacking, Sperber, Premack and Premack1995) notion of ‘looping effects’. An account that distributes agency evenly across actors, devices, and networks cannot, he argues, explain how representations intervene, because intervention runs through beings who grasp the world and revise their conduct accordingly. His realist alternative, he contends, supplies what our framework lacks.

What we agree on

Curran articulated representation as intervention before we did, and the semantic closeness is real. And many others did it before him, especially Callon. We are happy to differentiate the two terms, and the difference is substantive rather than terminological. ‘Representation as intervention’ names a thesis: that representations of social life act on social life. ‘Representational interventions’ names a mode within a wider taxonomy, denoting one of the ways performative processes are carried, organized, and made comparable across cases. Curran (Reference Curran2018) offers a theory of how a certain kind of representation works. We offer a map on which his theory occupies a definite and important location (Unal et al., Reference Unal, Polillo, Caliskan and MacKenzie2025). The relationship is not rivalry but nesting. We concede too that reflexivity matters and that representation deserves a theory rather than a gesture. The fair version of his third objection is that we lean on the word while leaving its workings implicit. What follows is the account we should have made explicit.

On the reading of Callon

Curran’s novelty objection leans on a specific claim: that one of the precursors we cite, namely Callon (Reference Callon2006), explicitly rejects the representational idiom, causing representation to enter the actor-network theory tradition only belatedly and, with our paper, in a somewhat ad hoc manner. We think this reading mistakes the target of Callon’s rejection. The phrase ‘representational idiom’ is not Callon’s own coinage. He takes it over from Pickering (Reference Pickering1995), and it has a precise and narrow content. What it names is the doctrine that the purpose of science is to produce representations that mirror a reality already fully formed and sitting independently of any description of it. That is representationalism understood as correspondence. It is this conception, representation as mirroring of a pre-given object, that Callon sets aside, and it is this very conception that Curran himself rejects when he insists that representation must not be reduced to mirroring. The two are therefore refusing the same thing. Callon does not abandon reference, description, or aboutness; he relocates them. To read his rejection of mirroring as a rejection of representation as such is to attribute to him a position he does not hold, and then to credit our paper with repairing a problem that was never there.

What Callon retains is in fact an articulated theory of reference, and this matters directly for Curran’s third critique. Statements in his account continue to point, to refer, and to be about their worlds: his singular existential statements designate entities located in time and space, and they carry what he calls ‘shifters’, the marks by which a statement indicates the situation it is about. Treatments of Black-Scholes (Callon, Reference Callon2006; MacKenzie, Reference MacKenzie2006) turn on exactly the directedness that Curran says the tradition neglects: the formula is about the securities practitioners construct and trade, which is why its world can be progressively realized. Callon even says, of that realized world, that the formula then describes and represents it correctly. What changes is not whether statements are about their objects, but the standard by which that aboutness is assessed: success and felicity, the mutual adjustment of statement and world, rather than correspondence to an object given in advance. Aboutness is theorized and distributed across the arrangement, not denied.

This also deflates the priority claim that is central to Curran’s reply. Hacking’s (Reference Hacking1983) division between science as representing and science as intervening, which he presents as his own re-appropriation, is already doing structural work in Callon’s Reference Callon2006 essay, which stages those two camps then sides with representing-through-intervening. Curran has elaborated that relation valuably, but it did not originate with him, and the tradition he describes as representation-averse was theorizing it all along. The disagreement that survives is not about whether aboutness exists in the social studies of finance (SSF) account, but about where it is located – anchored in human intentionality or distributed across a socio-technical agencement – and that is the ontological question our framework seeks to hold open rather than settle once and for all.

Where the heuristic charge misreads us

Curran is right that no taxonomy is innocent of presuppositions, and we withdraw any suggestion that DARN floats free of such commitments. But there is a difference between a framework that carries minimal commitments and one that advances a substantive theory of what representation is and what exists. DARN is ontologically agnostic in the sense that matters: it does not adjudicate between realist and anti-realist accounts of representation, and it is designed to be usable by scholars as well as designers (economic or not) who disagree about that adjudication (Caliskan and Wade, Reference Caliskan and Wade2022a; Reference Caliskan and Wade2022b). This is why we offer no critique of existing approaches. The point of a heuristic is to let seemingly incompatible substantive theories be compared side by side, which cannot be done from inside any one of them.

The statements Curran flags are not the metaphysical claims he takes them to be. When we say devices, actors, and networks possess agency, but only representations can be performative, we are drawing a boundary around what the word ‘performativity’ picks out, not making a claim about the making of the world. ‘Agency’, in the distributed sense long established in this tradition, means making a difference to an outcome, and many things do. ‘Performativity’ is the narrower relation in which a representation reshapes the situation it describes; ‘devices’ and ‘networks’ make differences constantly without describing anything. Reserving ‘performativity’ for representations is an analytic stipulation that buys clarity, and one can reject the stipulation but still use the map. The objection regarding the four categories of DARN rests on a misreading we should have prevented. Curran takes us to claim that representations of devices, actors, and networks are performative, then asks why only these four of the many others of things represented. DARN’s four terms are not a list of things that get represented. They are the modes through which representational interventions operate. A representation becomes performative by working through devices, through actors, through networks (or structures and infrastructures), or through its own circulation as representation. The question is not why represent networks rather than structures, but through what a representation acts. Structures are not excluded; they are redescribed as durable patterning of networks and devices.

Our aim is not to explain the forces that create agencements, where human intentionality might be central, but to theorize the performation process through which representations become performative of DARNs. The agencement is the mediator that supplies the felicity conditions for that outcome: Black-Scholes theory as a representation moves through an agencement of practitioners, regulators, market devices, and other entities, to financial prices as performed outputs. Our concern is with that process rather than with ranking the forms of agency. The four DARN elements are not arbitrary; they emerge from our review of the empirical literature as analytically distinct modes through which performativity operates. We treat them as performed outputs rather than as inherently performative, which does not deny them a role: as outputs of earlier struggles, they can become components of new agencements and help make other representations performative elsewhere, functioning as the ‘agencing actors’ that Diedrich et al. (Reference Diedrich, Cochoy and Guzman2023) describe.

The account of representation we owe him

Here we make the positive case, from DARN rather than from any single precursor. A representation, in our framework, is performative when it loops into devices, actors, and networks, and alters them, such that the social world comes to answer to the representation that describes it. Aboutness is not absent from this picture. A pricing model is about a security; a category is about the people it sorts. What DARN adds is that this aboutness is achieved and sustained through arrangements rather than secured in advance by an individual mind. The model is about a security because an infrastructure of data feeds, screens, settlement systems, and trained practitioners holds it in place. Remove the arrangement and the aboutness begins to come apart. Its referential dimension is distributed across the very modes our taxonomy names.

This lets us absorb Curran’s strongest example rather than oppose it. Hacking’s looping effects describe what happens when a representation intervenes on actors who are reflexive: for example, people grasp how they are classified and revise their conduct, which changes the classification, which changes conduct again (Hacking, Reference Hacking, Sperber, Premack and Premack1995). This is not a phenomenon our framework misses. It is one our framework locates. Looping is representational intervention running through the Actor mode, the special case in which the targets are conscious humans capable of taking themselves as objects of reflection. Curran’s paper gives us the sharpest available statement of how that mode works, and we should have cited him to say so.

The disagreement that remains is real but narrower than Curran’s framing suggests, and it runs in our favor on generality. Curran wants aboutness anchored in human intentionality and the reflexivity of conscious humans, and wants that to ground a general account of representation. Our objection is not that intentionality is unimportant. It is that intentionality cannot ground the general account, because a large and growing share of performative processes operate through devices and networks where no conscious aboutness is present at the point of action. An algorithmic pricing system performs a market without grasping anything. A credit infrastructure sorts populations through millions of automated decisions in which no one is reflecting on anyone. In the agentic and platform settings that increasingly define economic life, these are becoming the typical cases rather than marginal ones. Make human reflexivity the ground of representation, and you have a powerful theory of looping and no general theory of performativity, because the cases that matter most now delegate reflexive aboutness to devices and disperse it through networks. DARN does not deny reflexivity its place; it declines to let one mode stand for all four. Agency and performation are distributed, and for all. In other words, performative agency belongs to no single conscious agent but is shared across all four modes.